Multicore offloading of network processing

Abstract

A method for reassigning flows to cores in a multi-core network device includes receiving a packet flow and periodically determining a packet rate of the flow and the processing load on each of the worker cores. Unassigned flows are assigned to the least loaded core. If an assigned flow has a packet rate that exceeds a particular threshold proportion of the processing capacity of the currently assigned worker core, reassigning the flow to the lowest loaded worker core unless the resulting load would exceed the current load on the currently assigned worker core.

What is claimed is: 
     
       1. A method for processing network flows in a multi-core network device, the method comprising:
 receiving a packet flow at the multi-core network device having a plurality of worker cores; 
 periodically determining a packet rate of the packet flow and a processing load on each of the plurality of worker cores; 
 determining if the packet flow is currently assigned to one of the plurality of worker cores; 
 in response to determining that the packet flow is not assigned, assigning the packet flow to a worker core that has the lowest processing load among the plurality of worker cores; and 
 in response to determining that the packet flow is currently assigned and that the determined packet rate exceeds a particular threshold of the processing capacity of the currently assigned worker core:
 determining which of the plurality of worker cores has the lowest load; 
 determining if reallocating the packet flow to the lowest loaded worker core would cause the load on the lowest loaded worker core to exceed the current load on the currently assigned worker core; and 
 in response to determining that reallocating the packet flow would not cause the load on the lowest loaded worker core to exceed the current load on the currently assigned worker core, reallocating the packet flow to the lowest loaded worker core. 
 
 
     
     
       2. The method of  claim 1  wherein the multi-core network device is programmed and configured to operate using a modular software architecture, the method further comprising:
 generating an instance of an offloading module adapted for processing a particular class of flow; and 
 inserting the instance into a chain of modules in the modular software architecture, wherein the offloading module performs said reallocating of the packet flow. 
 
     
     
       3. The method of  claim 2  wherein the instance of the offloading module is inserted such that an input of the module is connected between an output of a forwarding information base (FIB) module and an input to a module configured for processing the particular class of flow. 
     
     
       4. The method of  claim 2  wherein the particular class of flow is at least one of an IP security (ipsec) decryption or ipsec encryption flow. 
     
     
       5. The method of  claim 4  wherein the particular class of flow is restricted to non-control-plane flows. 
     
     
       6. The method of  claim 2  wherein the modular software architecture is based upon a Berkeley Extensible Software Switch (BESS). 
     
     
       7. The method of  claim 1  wherein the particular threshold of the processing capacity is between 50 and 60 percent. 
     
     
       8. The method of  claim 1  wherein the determining the packet rate of the packet flow and the processing load on each of the worker cores periodically occurs between every 10 to 60 seconds. 
     
     
       9. The method of  claim 1  wherein the reallocating the packet flow to the lowest loaded worker core is further in response to estimating that a relative difference in loads between the lowest loaded worker core and the currently assigned core is reduced by at least a minimum threshold. 
     
     
       10. The method of  claim 9  wherein the minimum threshold is six percent of processing capacity. 
     
     
       11. A system for processing network flows in a multi-core network device, the system comprising:
 one or more processors programmed and configured to perform:
 receiving a packet flow at the multi-core network device having a plurality of worker cores; 
 periodically determining a packet rate of the packet flow and a processing load on each of the plurality of worker cores; 
 determining if the packet flow is currently assigned to one of the plurality of worker cores; 
 in response to determining that the packet flow is not assigned, assigning the packet flow to a worker core that has the lowest processing load among the plurality of worker cores; and 
 in response to determining that the packet flow is currently assigned and that the determined packet rate exceeds a particular threshold of the processing capacity of the currently assigned worker core:
 determining which of the plurality of worker cores has the lowest load; 
 determining if reallocating the packet flow to the lowest loaded worker core would cause the load on the lowest loaded worker core to exceed the current load on the currently assigned worker core; and 
 in response to determining that reallocating the packet flow would not cause the load on the lowest loaded worker core to exceed the current load on the currently assigned worker core, reallocating the packet flow to the lowest loaded worker core. 
 
 
 
     
     
       12. The system of  claim 11  wherein the one or more processors are further programmed and configured with a modular software architecture to:
 generate an instance of an offloading module adapted for processing a particular class of flow; and 
 insert the instance into a chain of modules in the modular software architecture, wherein the offloading module performs the reallocating of the packet flow. 
 
     
     
       13. The system of  claim 12  wherein the instance of the offloading module is inserted such that an input of the module is connected between an output of a forwarding information base (FIB) module and an input to a module configured for processing the particular class of flow. 
     
     
       14. The system of  claim 12  wherein the particular class of flow is at least one of an IP security (ipsec) decryption or ipsec encryption flow. 
     
     
       15. The system of  claim 14  wherein the particular class of flow is restricted to non-control-plane flows. 
     
     
       16. The system of  claim 12  wherein the modular software architecture is based upon a Berkeley Extensible Software Switch (BESS). 
     
     
       17. The system of  claim 11  wherein the particular threshold of the processing capacity is between 50 and 60 percent. 
     
     
       18. The system of  claim 11  wherein the determining the packet rate of the packet flow and the processing load on each of the worker cores periodically occurs every 10 to 60 seconds. 
     
     
       19. The system of  claim 11  wherein the reallocating the packet flow to the lowest loaded worker core is further in response to estimating that a relative difference in loads between the lowest loaded worker core and the currently assigned core is reduced by at least a minimum threshold. 
     
     
       20. The system of  claim 19  wherein the minimum threshold is six percent of processing capacity.